EDock-ML: A web server for using ensemble docking with machine learning to aid drug discovery.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Receptor flexibility is crucial in molecular docking for drug discovery.
• Ensemble docking offers an efficient method to address receptor flexibility.
• Machine learning can enhance the interpretation of docking scores for predicting compound efficacy.

Purpose of the Study:

• To introduce EDock-ML, a web server integrating ensemble docking and machine learning.
• To provide a user-friendly tool for assessing the potential utility of compounds in drug discovery.
• To improve the accuracy of predicting compound activity through protein-specific machine learning models.

Main Methods:

• Utilizes ensemble docking to simulate receptor flexibility.
• Employs machine learning models trained on a protein-by-protein basis.
• Accepts compound input via ZINC database IDs or chemical drawing files.
• Provides output to guide decisions on compound progression in drug discovery.

Main Results:

• EDock-ML facilitates the decision-making process for drug candidates.
• The server's approach allows for direct use by novice researchers without parameter tuning.
• Machine learning models are developed incrementally to enhance prediction accuracy for specific proteins.

Conclusions:

• EDock-ML offers a valuable resource for drug discovery by combining ensemble docking and machine learning.
• The web server simplifies the evaluation of compound activity, aiding researchers in prioritizing candidates.
• The protein-specific, bottom-up machine learning approach enhances predictive power and usability.